Use of Arthropod Rarity for Area Prioritisation: Insights from the Azorean Islands

Simone Fattorini1,2*, Pedro Cardoso1,3, Franc¸ois Rigal1, Paulo A. V. Borges1 1 Azorean Biodiversity Group, Universidade dos Ac¸ores, Departamento de Cieˆncias Agra´rias CITA-A, Pico da Urze, Angra do Heroı´smo, Portugal, 2 Water Ecology Team, Department of Biotechnology and Biosciences, University of Milano Bicocca, Piazza della Scienza 2, Milan, Italy, 3 Smithsonian Institution, National Museum of Natural History, Washington, D.C., United States of America

Abstract We investigated the conservation concern of Azorean forest fragments and the entire Terceira Island surface using arthropod species vulnerability as defined by the Kattan index, which is based on species rarity. Species rarity was evaluated according to geographical distribution (endemic vs. non endemic species), habitat specialization (distribution across biotopes) and population size (individuals collected in standardized samples). Geographical rarity was considered at ‘global’ scale (species endemic to the Azorean islands) and ‘regional’ scale (single island endemics). Measures of species vulnerability were combined into two indices of conservation concern for each forest fragment: (1) the Biodiversity Conservation Concern index, BCC, which reflects the average rarity score of the species present in a site, and (2) one proposed here and termed Biodiversity Conservation Weight, BCW, which reflects the sum of rarity scores of the same species assemblage. BCW was preferable to prioritise the areas with highest number of vulnerable species, whereas BCC helped the identification of areas with few, but highly threatened species due to a combination of different types of rarity. A novel approach is introduced in which BCC and BCW indices were also adapted to deal with probabilities of occurrence instead of presence/absence data. The new probabilistic indices, termed pBCC and pBCW, were applied to Terceira Island for which we modelled species distributions to reconstruct species occurrence with different degree of probability also in areas from which data were not available. The application of the probabilistic indices revealed that some island sectors occupied by secondary vegetation, and hence not included in the current set of protected areas, may in fact host some rare species. This result suggests that protecting marginal non-natural areas which are however reservoirs of vulnerable species may also be important, especially when areas with well preserved primary habitats are scarce.

Citation: Fattorini S, Cardoso P, Rigal F, Borges PAV (2012) Use of Arthropod Rarity for Area Prioritisation: Insights from the Azorean Islands. PLoS ONE 7(3): e33995. doi:10.1371/journal.pone.0033995 Editor: Brock Fenton, University of Western Ontario, Canada Received December 15, 2011; Accepted February 22, 2012; Published March 30, 2012 Copyright: ß 2012 Fattorini et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: Data used in this manuscript were obtained in the projects funded by Direcçaõ Regional dos Recursos Florestais (Project: 17.01-080203, 1999–2004) and Direcçaõ Regional da Cieˆncia e Tecnologia (Project: ‘‘Consequences of land-use change on Azorean fauna and flora - the 2010 Target, M.2.1.2/I/003/2008). Grants and fellowships to the authors were provided by FundaçaõCieˆncia e Tecnologia (SFRH/BPD/40688/2007 and PTDC/BIA-BEC/100182/2008 for PC and FR, respectively) and Azorean Biodiversity Group (CITA-A) (Summer 2010 and Summer 2011 grants to SF). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail: [email protected]

Introduction identification of biodiversity hotspots and the selection of priority areas are still generally based on the occurrence of target species Protected areas are considered one of the most effective and among vertebrates and vascular plants [20]. This contrasts with cost-efficient ways to conserve habitats and viable populations of the fact that invertebrates, and in particular arthropods, are the species, representative of the biological diversity of the Earth [1,2]. most diverse and abundant animal group in virtually all biotopes, However, also the landscape outside reserves could have an performing a number of ecosystem functions that are irreplaceable important, albeit usually overlooked, role in the conservation of [21], and include the vast majority of species threatened by particular species [3–6]. Selection of priority areas for biological extinction [22]. In general, it is assumed that invertebrates are too conservation has been long driven by sociological, economical, poorly known for driving conservation decisions [20]. This is due and practical reasons, sometimes with tenuous scientific support to a number of impediments, namely the scarce or non-existent [7,8]. Recent developments in systematic conservation planning knowledge about most species [21], including about their have put forward the need for more scientifically well-founded distribution (the Wallacean shortfall [23]), changes in space and criteria for area prioritisation [7–13]. Obviously, one of the most time (the Prestonian shortfall [21]) and vulnerability to habitat commonly used criteria for locating areas of conservation concern change (the Hutchinsonian shortfall [24]). is the presence of target species [14,15,16] or biotopes [17]. A variety of species distribution modelling techniques [25] has However, preserving umbrella or indicator species does not recently been developed, and their application in conservation necessarily coincide with preserving the biodiversity at large and planning has been advocated [26], allowing a possible practical protected areas established for conserving certain target species do solution for the Wallacean shortfall [21]. In addition, recent work not protect automatically all imperilled species [18,19]. The showed that it is relatively easy to obtain reliable measures of

PLoS ONE | www.plosone.org 1 March 2012 | Volume 7 | Issue 3 | e33995 Arthropod Rarity and Hotspot Identification species threat status and conservation value for arthropods, similar Europe and 2150 km east of the North American continent. The to those used for plants and mammals [21,27,28], which proved native forest in the Azores is characterized by an association of important to partly overcome the Prestonian and Hutchinsonian native (many endemic) evergreen shrub and tree species. shortfalls [21]. Commonly known as Laurisilva, this forest occupied most of the The Azorean Islands, a remote archipelago in the Atlantic surface of all the islands before human settlement almost 600 years Ocean, offer the unique opportunity of exploring the contribution ago. However, native forests are now mostly restricted to high and that arthropods can offer to the identification of rarity hotspots. In steep areas, while most of the islands are covered by exotic these islands, the vast majority of endemic species are arthropods plantations of Cryptomeria japonica and Eucalyptus spp., abandoned [29]. Yet, selection of priority areas for conservation on these islands fields now dominated by Pittosporum undulatum, semi-natural has been mainly driven by biotopes, rare vascular plants and a few pastures, and intensively managed pastures. Although protected vertebrates [30,31]. In this paper, we took advantage of data native forest covers less than 3% of the total area of the collected during a long term project of arthropod inventorying and archipelago, it is the biotope in which the great majority of the monitoring [32–35], in order (1) to develop a multidimensional endemic plant and animal species occur in the Azores characterization of arthropod species rarity based on standardized [32,33,35,36]. sampling; (2) to use rarity measures to derive indices of species In this study, we first considered 18 native forest fragments vulnerability to extinction; and (3) to use such indices to classify distributed across seven islands of the archipelago: Santa Maria, areas according to the vulnerability of the species they harbour. This Saõ Miguel, Terceira, Saõ Jorge, Faial, Pico and Flores (Table 1, was done for (1) all native forest fragments in all the Azorean islands see [36] for details). This corresponds to most of the native forest and (2) all areas, irrespectively of their biotope type, in Terceira, the extent of the Azores. All these areas are now protected under island with the most comprehensive data. different regimes [37]. In each forest fragment, arthropod sampling was conducted Methods using the same standardized protocols to collect both ground dwelling arthropods (by pitfall traps) and canopy arthropods (by We employed a multistep modelling approach to identify beating). Using the same sampling protocol we also collected priority areas for conservation. In the following sections, we individuals across six different land uses (i.e. high altitude natural describe the main points of our methodological framework. grasslands, peat bogs, exotic forests, semi-natural pastures, Further details about the analyses are given in Information S1. intensively managed pastures, canopies of orchards) for six islands: Santa Maria, Terceira, Saõ Jorge, Faial, Pico, and Flores (see Study areas and sampling [33,38,39] and Information S1). The Azores archipelago stretches out over 615 km in the North All necessary permits from the Azorean Nature Parks for each Atlantic Ocean (37–40uN, 25–31uW), 1584 km west of southern of the studied island were obtained for the described field studies.

Table 1. IUCN levels of protection (according to [37]), Index of Biodiversity Conservation Concern, and Index of Biodiversity Conservation Weight for 18 native forest fragments on the Azorean Islands.

IUCN levels of Forest Fragment (island) protection BCC with SIE BCC with AZE BCW with SIE BCW with AZE

Atalhada (S. Miguel) IV 0.132 0.329 0.145 0.248 Biscoito da Ferraria (Terceira) I 0.153 0.375 0.186 0.311 Cabec¸o do Fogo (Faial) IV 0.111 0.259 0.094 0.153 Caldeira do Faial (Faial) I 0.122 0.332 0.096 0.181 Caldeira Guilherme Moniz (Terceira) VI 0.094 0.252 0.091 0.167 Caldeiras Funda e Rasa (Flores) I 0.111 0.329 0.113 0.231 Caveiro (Pico) I 0.110 0.350 0.108 0.237 Graminhais (S. Miguel) IV 0.090 0.329 0.071 0.179 Lagoa do Caiado (Pico) IV 0.094 0.337 0.086 0.212 Miste´rio da Prainha (Pico) I 0.094 0.300 0.127 0.276 Morro Alto e Pico da Se´ (Flores) I 0.132 0.335 0.148 0.256 Pico Alto (Sta Maria) IV 0.155 0.348 0.185 0.284 Pico da Vara (S. Miguel) I 0.119 0.305 0.147 0.257 Pico do Galhardo (Terceira) IV 0.102 0.316 0.112 0.238 Pico Pinheiro (S. Jorge) IV 0.129 0.350 0.137 0.255 Serra Sta. Ba´rbara (Terceira) I 0.157 0.374 0.219 0.354 Terra Brava (Terceira) I 0.110 0.326 0.138 0.279 Topo (S. Jorge) V 0.128 0.385 0.124 0.256

IUCN levels of protection: I –Natural Reserve; III –Natural Monument; IV –Habitat and species management; V –Protected Landscape; VI –Resources management. BCC: Index of Biodiversity Conservation Concern; BCW: Index of Biodiversity Conservation Weight. BCC and BCW were calculated using single island endemics (SIEs) and Azorean endemics (AZEs). doi:10.1371/journal.pone.0033995.t001

PLoS ONE | www.plosone.org 2 March 2012 | Volume 7 | Issue 3 | e33995 Arthropod Rarity and Hotspot Identification

None of the species sampled are protected by Azorean, Portuguese measures indicate that the index is robust to different ways of or International laws. However, this sampling allowed us to inform calculating species rarity. the Azorean Government about the distribution of restricted endemic species for improving the design of current Protected Forest fragment ranking Areas (Borges et al., unpublished Reports). We ranked forest fragments according to two different measures of prioritisation. Measures of species rarity - We used the Biodiversity Conservation Concern (BCC) index In order to fulfil Hartley and Kunin’s recommendations of [56] whose original formulation was modified to make it more considering different aspects of rarity [40], species rarity was general, as observed in [57]. With the new formulation, BCC can assessed here using a multidimensional characterization that takes be calculated as: into account: (1) geographical distribution (wide/narrow distribu- PL (ai{amin) tion), (2) abundance (abundant/scarce population), and (3) habitat ~ BCC~ i 1 ð1Þ specificity (low/high habitat specificity) [41,42]. Such a multidi- L(amax{amin) mensional characterization of species rarity has been successfully applied to vertebrates [43,44,45], arthropods [46–49] and where L is the local species richness, ai is the vulnerability index bryophytes [50]. assigned to the ith species (as defined above), amin is the minimum weight among all species; and amax is maximum weight among all Geographical distribution species. Estimating the geographical rarity of a species depends on the The BCC index has been previously applied to identify priority spatial scale of analysis [51,52], so we adopted a two-level areas or biotopes for butterflies in Mediterranean islands and approach. At a global level, we considered as geographically rare European countries [56–59], fish in France [60], tenebrionids, the species which are endemic to the Azorean Islands, even if butterflies, birds and mammals in the Central Apennines [47,48]. distributed in more than one island (hereafter AZE species). At a - We also used a new index, the Biodiversity Conservation regional level, we considered as geographically rare the species Weight (BCW) index, also based on species vulnerability. The BCC which are endemic to single Azorean Islands (singe island index is a ‘relative measure’, which means that it is not sensitive to endemics, hereafter SIEs). Endemics are typically considered as species richness. This may be an advantage to compare species taxa of conservation concern [53,54], and this approach also assemblages with different species richness [48,56], but poses some ensures that endemic taxa are scored as important, at least in problems. For example, an assemblage with a single species, having terms of geographical rarity, from a global and a regional this species amax, would receive the same score as an assemblage perspective. with 10 species, all with amax. Or worse, an assemblage with a single species with amax has a higher score than an assemblage with 10 Abundance species, 9 with amax and one with ai,amax. To overcome this problem, we have calculated the BCW as follows: To calculate the relative abundance of each species in the Azores we used all the standardized transects available for all main biotopes in seven of the nine islands (Corvo and Graciosa were not PL (ai{amin) sampled since they have entirely lost their native forest; see more ~ BCW~ i 1 ð2Þ details in [33,35,39], and Information S1). Species with abundance PS below the median were classified as rare. (ai{amin) i~1

Habitat specificity where S is the total species richness for all sites (all other symbols as We used species abundances across the biotopes occurring on for BCC, see above). the study islands to calculate species habitat specificity using the Spearman rank correlations were used to test inter-correlations Shannon H9 index [55]. Species with H9 values below the median between BCC and BCW values. were classified as rare [43]. Potential distribution modelling Vulnerability index In many cases the features to rank are not discrete, relatively Species with smaller ranges, lower abundances and narrower large, units for which most existing species are known, such as the biotope ranges tend to experience higher levels of threat [45]. 18 Azorean forest fragments. Especially for arthropods and other Thus, using species categorisation into the rarity forms described small organisms, just because a species is not known from a site above (i.e. geographic distribution, abundance and habitat does not mean it is not present. Often it was just not searched for specificity), we calculated an index of species vulnerability as or not found and such site can be overlooked in conservation proposed by Kattan [43]. priority exercises. We calculated two measures of the Kattan index, considering Thus, for Terceira, the island for which more information was alternatively as geographically rare only SIEs or all the AZEs. x2- available, we calculated and mapped potential BCC and BCW tests were used to determine the independence of the three (pBCC and pBCW) based on probabilistic species distributions. measures of rarity [43]. For this, we used the maximum entropy algorithm [61,62] to Spearman rank correlations were used to test inter-correlations model species distributions on this island using climatic data, among number of islands from which a species is known (NISL), landscape maps and topographical and geographical information number of biotopes occupied by a species (NBIO), H9 measure of [63–65] (see Information S1 for details). habitat specificity, species abundance, and Kattan indices. NISL and NBIO were considered as measures of geographical rarity and Mapping of potential rarity habitat specificity alternative to those used to construct the Kattan The BCC and BCW indices were designed to deal with index. Correlations between the Kattan index and these two occurrence data, not with probabilities of occurrence. One

PLoS ONE | www.plosone.org 3 March 2012 | Volume 7 | Issue 3 | e33995 Arthropod Rarity and Hotspot Identification possibility to use them with the latter type of data would be to Forest fragment ranking convert probability maps into presence/absence maps by using a Values of Index of Biodiversity Conservation Concern (BCC) threshold in probabilities above which the species would be and Index of Biodiversity Conservation Weight (BCW) are considered to be present [66]. This would however cause three reported in Table 1, and their intercorrelations in Information S5. shortcomings. Firstly, the best threshold is hard to define, although Although correlation values between indexes varied, the a few guidelines exist [66,67]. Secondly, this would imply a loss of following fragments were consistently placed in the third quartiles information. Thirdly, this would consider as completely different for all four indices (BCC and BCW using SIEs and AZEs): Serra some sites with very similar species composition if such sites were Sta. Ba´rbara, Biscoito da Ferraria (the two largest fragments in very close to the threshold for one or a few rare species. Terceira) and Pico Alto (the only fragment in the oldest island, Thus we preferred to use modified versions of the BCC and Santa Maria). BCW formulas to explicitly cope with probabilities of occurrence Focusing on the top five ranked fragments (third quartile) for (see Supplementary Information about Methods for details). The each index (Table 2), the five fragments selected by the BCC and formulas for potential BCC (pBCC) and potential BCW (pBCW) BCW with SIEs captured about 80% of the entire species richness are therefore: of all 18 fragments. Species captured by these two indices showed also relatively high mean values for vulnerability indices (Table 2). PS pOi(ai{amin) Mapping of potential rarity ~ pBCC~ i 1 3 All 47 species probability maps had AUC values above 0.7 and { ð Þ pS(amax amin) we considered them as reliable (Information S6). Highest values of and potential species richness (Fig. 2) were concentrated in the five forest fragments of Terceira: Serra Sta. Ba´rbara, Biscoito da Ferraria, Terra Brava, Pico Galhardo and Caldeira de Guilherme Moniz. PS Use of pBCC with Azorean endemics produced a somewhat pOi(ai{amin) ~ similar pattern (Fig. 3B), while the use of only SIEs as geographically pBCW~ i 1 ð4Þ PS rare species highlighted a more complex pattern (Fig. 3A). This { (ai amin) more restrictive SIE approach, more than for the aforementioned ~ i 1 areas, gave relatively high scores to the protected areas of Monte where for each cell: S is the total species richness for all sites; pS is Brasil (southernmost tip of the island) and Serreta (northeastern PS Terceira), in the coastal areas of the island. The pBCC with SIE also the potential species richness (pS = pOi), pOi is the probability of highlighted an important patch in the southwestern part of the i~1 island (Fonte do Bastardo). Use of pBCW (Fig. 3 C and D) gave occurrence of species i, ai is the weight of species i; amin is the results somewhat similar to those achieved using potential species minimum weight among all species and amax is the maximum richness or pBCC with Azorean endemics, although even more weight among all species. strongly emphasizing the importance of native forest fragments. To emphasize differences in the outputs of pBCC and pBCW, we rescaled previous maps from 0 to 1 and did a simple Results subtraction of pBCC from pBCW (Fig. 4). This shows that pBCC In total, we considered 219 arthropod species, 178 of which are is giving more importance to low altitude areas, most notably found in the 18 studied protected areas. Of these 178 species, 82 Monte Brasil, while pBCW is giving more importance to native are considered Azorean endemics (AZE) and of those 26 are Single forests or high altitude areas. Island Endemics (SIEs) (see Information S2). Discussion Vulnerability index Rabinowitz’s approach to rarity Although non-rare species were the most abundant category Previous studies using Rabinowitz’s forms of rarity [41,42] (28–40% according to the measure of geographical rarity which is found that while a high proportion of species have relatively small used), a high proportion of species was rare for at least one geographical ranges, only few species are widespread and criterion (Fig. 1). Using the SIE criterion, about 5% of the species abundant, and the condition of ‘abundant and localized’ is were rare for all rarity dimensions (geography, abundance and extremely rare since locally abundant populations tend to rapidly habitat). This percentage increased substantially with the use of occupy new sites [51,52,68,69]. However, it is noteworthy to AZEs reaching close to 10%. consider the scale of analysis, and hence the way geographical 2 The results of the x tests indicate that the hypothesis of overall rarity is assessed. When considering as geographically rare only the independence of the three rarity dimensions is rejected (Informa- SIEs, we found a relatively small percentage (about 9%) of species tion S3). However, separate analyses of the 262 tables indicate which were abundant and geographically restricted. But this that distribution and abundance are jointly independent factors percentage was about 29% when endemics were considered as (Information S3). geographically restricted. That is, almost one third of the AZEs at Both Kattan indices were strongly correlated with the original the archipelago scale were considered abundant, which implies measures of species habitat specialization (H9) and abundances that many of the endemics that were able to occupy more than one from which the indexes have been obtained (Information S4). island were also successful in building large populations in most Interestingly, both Kattan indices were also correlated with the islands. Because the Kattan index used as a vulnerability measure number of biotopes a species occupies and the number of islands in the BCC and BCW indexes gives more weight to geographical from which a species is known, which can be considered rarity, it is critical to carefully consider the scale of analysis. alternative measures of habitat specialization and geographical Moreover, comparisons of multiple taxa within the same rarity (Information S4). geographical context revealed that proportions of different

PLoS ONE | www.plosone.org 4 March 2012 | Volume 7 | Issue 3 | e33995 Arthropod Rarity and Hotspot Identification

Figure 1. Percentages of the seven categories of arthropod rarity. A total of 178 arthropod species in 18 forest fragments in the Azorean Islands were considered with different criteria for endemics: (A) only single island endemics (SIEs) were considered geographically rare; (B) all Azorean endemics (AZEs) were considered geographically rare. doi:10.1371/journal.pone.0033995.g001 categories of rarity tend to change considerably among taxa [48]. habitat specificity is measured. This qualifies the Kattan index as a Thus, no generalization seems possible and rarity measures always good synthetic measure of species ‘rarity’. have a relative value, depending on the particular assemblage of species under study (cf. also [43,44,45]). Prioritisations of biotopes and areas (BCC vs. BCW) On the other hand, the Kattan index was very efficient in Although species are the primary target of conservation efforts, summarizing the three dimensions of rarity and it was also proven a number of impediments, including the Linnean shortfall to be robust to variations in the way geographical rarity and (incomplete taxonomic knowledge), the Wallacean shortfall

Table 2. Number (and percentages) of species included in the first five ranked fragments according to Index of Biodiversity Conservation Concern and Index of Biodiversity Conservation Weight, with indication of Mean (and Standard Deviation) values of vulnerability (Kattan index) of the species included in the selected fragments.

BCC with SIE BCC with AZE BCW with SIE BCW with AZE

Captured species richness (%) 141 (79.2) 115 (64.6) 139 (78.1) 130 (73.0) Mean (SD) value of Kattan index of included 2.454 (1.830) 2.409 (1.910) 2.511 (1.931) 2.377 (1.814) species with SIE criterion Mean (SD) value of Kattan index of included 3.624 (2.316) 3.765 (2.313) 3.698 (2.370) 3.638 (2.353) species with AZE criterion

BCC: Index of Biodiversity Conservation Concern; BCW: Index of Biodiversity Conservation Weight. BCC and BCW were calculated using single island endemics (SIEs) and Azorean endemics (AZEs) as alternative criteria for geographical rarity. doi:10.1371/journal.pone.0033995.t002

PLoS ONE | www.plosone.org 5 March 2012 | Volume 7 | Issue 3 | e33995 Arthropod Rarity and Hotspot Identification

and biology is limited, thus surpassing the aforementioned shortfalls. Also, their combined use in the Kattan index may be particularly useful to obtain a general evaluation of species vulnerability. After a large number of species are evaluated, their vulnerability can be used to identify priority areas. In this study, we used two indices based on species vulnerability, the Biodiversity Conservation Concern (BCC, introduced by [56]) and the Biodiversity Conservation Weight (BCW) (introduced here) to prioritise forest fragments. The results provided by these two indices generally differ. BCC places more emphasis on species-poor areas which may contain, however, high proportions of mostly vulnerable species, whereas BCW tends to identify areas which have large numbers of highly vulnerable species. Although the BCW may appear to give a more logical signal, BCC can be used to drive attention to areas with few, but very rare threatened species. This can be important for areas occupied by biotopes Figure 2. Potential arthropod species richness on Terceira which host few, but highly specialized species, such as high altitude Island. Species richness is based on probability of occurrence. Colder colours (dark blue) represent low values (minimum value = 4.055) and open biotopes [48] or caves [73]. For the best preserved areas, the hot colours (red) represent high values (maximum value = 29.251). The two indices tend to give similar prioritisations, but the BCC tends theoretical range is 0–47 as 47 species were evaluated. to emphasize degraded areas which still host few imperilled doi:10.1371/journal.pone.0033995.g002 species. This calls attention for the need to create additional measures of conservation management to non-natural areas (incomplete information on species distribution), the Prestonian [3,4,5]. In small territories like islands in which the matrix shortfall (lack of adequate estimates of population abundance and surrounding the protected areas concentrates most of the intensive changes in space and time) and the Hutchinsonian shortfall forest and agriculture activities, those species located in isolated (incomplete knowledge of species relationships with the environ- pockets are in high danger of extinction. ment) [21] make generally impractical the adoption of species- It is noteworthy that the BCC and BCW indices tend to give the focused actions (e.g. action plans) for arthropods. Thus, arthropod highest values to the same fragments when using different criteria conservation is generally based mainly on the identification of of geographical rarity. However, the two indices may give different priority sites selected by the occurrence of priority species [70], results in less obvious cases, for example for fragments with few, assuming that preservation of the biotope of that/those species will but very vulnerable species. An important source of bias in the use automatically allow conservation of other imperilled species of these indices in locating priority areas may be the inadequate [71,72]. Rarity measures are widely recognized as good surrogates knowledge of species distribution (Wallacean shortfall). In of species extinction risk and can be obtained also when particular, failure to detect species in areas where they are in information on species taxonomy, distribution, population size fact present, can bias results in favour of the best sampled areas.

Figure 3. Maps of indices of arthropod conservation in Terceira. A and B illustrate potential Biodiversity Conservation Concern (pBCC). C and D illustrate potential Biodiversity Conservation Weight (pBCW). Colder colours represent low values and hot colours represent high values. Maps of figures A and C were calculated using only single island endemics (SIEs) as geographically rare species (ranges: 0.031–0.175 and 0.072–0.553, respectively). Maps of figures B and D were calculated using all Azorean endemics (AZEs) as geographically rare species (ranges: 0.081–0.282 and 0.051–0.638, respectively). doi:10.1371/journal.pone.0033995.g003

PLoS ONE | www.plosone.org 6 March 2012 | Volume 7 | Issue 3 | e33995 Arthropod Rarity and Hotspot Identification

conservation of Azorean soil epigean arthropod biodiversity [32,36]. In this current contribution, we evaluate the ability of different indices to reflect species assemblage importance, calculating the percentage of total richness included in the top ranked fragments for each metric (see also [75]). On the whole, the top five fragments included about 65–80% of total richness. The best results were obtained using BCC and BCW with SIEs. Thus, the use of SIEs seems to select areas which capture more species than those found using AZEs. This highlights the importance of native fragments that have unique species like the small and disturbed area of Pico Alto in Santa Maria (see also [32,39]). When ranking sites based on BCW, top native protected areas are mainly large pristine reserves, with exception of Pico Alto in Santa Maria. Pico Alto region is located in the archipelago’s oldest island and is a hotspot of biodiversity [32], in which over 57 endemic arthropod species are known, i.e. 21% of the Azorean endemic arthropods occur in an area representing ,0.25% of Azorean native forests. IUCN levels of protection for the Azorean native forests not always gave the higher priority to the most important areas. This is the case of Pico Alto (Santa Maria), Atalhada (Sa˜o Miguel) and Pico Pinheiro and Topo (Sa˜o Jorge) that score high in BCC – SIE or BCC – AZE, but have only a level of protection IV or V in the Azores (see Table 1). Most of these areas are highly disturbed [36], but still maintain important populations of unique species. This reveals the importance of considering not only a dual classification of protected/unprotected in spatial conservation planning, but to consider also the category of the protected areas and how well each category is able to guarantee the persistence of each species in the future. If some species are able to withstand some human intervention over their habitat, other may not and low protection categories may be insufficient. Figure 4. Differences between potential Biodiversity Conser- vation Concern (pBCC) and potential Biodiversity Conserva- tion Weight (pBCW). All maps were rescaled from 0 to 1 and pBCW Conclusions were subtracted from pBCC. Cold colours represent sites where pBCW is We used two indices to rank Azorean forest fragments and the higher than pBCC and hot colours represent sites where pBCC is higher entire area of Terceira Island according to arthropod species than pBCW. Values were calculated using only single island endemics vulnerability. To assess species vulnerability we referred to species (SIEs) (A) and all Azorean endemics (AZEs) (B) as geographically rare species. rarity. Species rarity was evaluated according to geographical doi:10.1371/journal.pone.0033995.g004 distribution, habitat specialization and population size of the species. Because geographical rarity can be assessed at different For example, comparing known patterns of Amazon plant scales, we performed our analyses considering two possible diversity with those reconstructed using modelled full distributions, classifications: at ‘global’ scale, we considered as rare the species Hopkins [74] showed that the ‘real’ diversity map of Amazonian endemic to the Azorean islands (AZEs); at ‘regional’ scale, only plant richness might be very different from the ‘known’ pattern. those endemic to single islands (SIEs). These alternative measures For this reason, in our study, we modelled potential arthropod of geographical rarity tend to produce different outcomes. We think no particular choice can be recommended in general, species distribution on Terceira Island, and then calculated for because it depends on the aim of the study. In our case, for each geographical unit the pBCC and pBCW indices on the basis example, the use of SIEs may be more appropriate to prioritise of the probability of occurrence of each species. This novel forest fragments among islands because it enhances the total approach allowed the identification of some areas that are number of species included in the final set of prioritised areas. potentially important for the conservation of biodiversity in Using synthetic indices to prioritise areas according to species Terceira Island, even if such areas were never sampled. For vulnerability also raises the problem whether applying an absolute example, the area of Monte Brasil, not included – and hence not or a relative measure, i.e. whether considering the overall weight evaluated – among the analysed forest fragments because occupied obtained by the sum of the vulnerability measures of the species by secondary vegetation, may also be important to preserve if the occurring in a given area (as in the BCW), or dividing this sum by objective is to guarantee the persistence of the endemic biota. species richness (as in the BCC). In general, an absolute index Some endemic species (in particular low altitude specialized seems preferable to prioritise the areas with the highest numbers of species, such as the endemic weevil Drouetius azoricus parallelirostris) vulnerable species, but a relative index may help the identification still occur in this area. of areas with few, but highly imperilled species. Thus, the two approaches should be used in tandem for a ‘balanced’ overview of Patterns of prioritisation for the Azorean native forest conservation priorities. Because areas are ranked on the basis of fragments the species they host, incomplete knowledge of species distributions We have previously examined the relative value of 18 forest can produce wrong prioritisations in favour of the best sampled fragments in seven of the Azorean islands to improve the areas. Moreover, common practice to rank areas in biological

PLoS ONE | www.plosone.org 7 March 2012 | Volume 7 | Issue 3 | e33995 Arthropod Rarity and Hotspot Identification conservation is to define a priori the areas to compare and then to Information S3 Results of the x2 tests of independence rank them according to the species. This might overlook important of the three dimensions of rarity for the Azorean areas which where not considered because of lack of data. Recent arthropods. development of procedures to model species distributions allows (PDF) the reconstruction of maps of species occurrence with different Information S4 Correlation (Spearman rank coefficient) be- degrees of probability covering areas from which data are not tween measures of rarity for 178 arthropods of the Azorean available. An application of such approach to the arthropods of Islands. Terceira revealed that some island sectors occupied by secondary vegetation, and hence not included among the areas analysed for Information S555 forest fragment prioritisation, may in fact host some vulnerable Correlation (Spearman rank coefficient) between Index of Biodiversity species. The natural landscapes of the Azorean Islands have been Conservation Concern and Index of Biodiversity Conservation Weight almost completely destroyed and primary forests are reduced to for 18 Azorean forest fragments. very few, sparse and small fragments. In such circumstance, (PDF) protecting non-natural areas which are however reservoirs of Information S6 Area Under the Curve (AUC) values for imperilled species may be also important [3,4,5]. the potential distribution modelling of all 47 studied species. Supporting Information (PDF)

Information S1 Detailed description of sampling proce- Aknowledgments dures and statistical calculation. (PDF) We are grateful to two anonymous reviewers for their constructive comments on a previous version of this paper. Information S2 Endemic status, number of occupied islands, number of occupied habitats, Shannon H9 index Author Contributions of habitat specificity and abundance of each arthropod species found in the Azorean forest fragments. Conceived and designed the experiments: SF PC FR PAVB. Analyzed the data: SF PC FR. Wrote the paper: SF PC FR PAVB. Collected the data: (XLS) PC PAVB.

References 1. Ervin J (2003) Protected area assessments in perspective. Bio Science 53: 20. Pereira HM, Cooper HD (2006) Towards the global monitoring of biodiversity 819–822. change. TREE 21: 123–129. 2. Chape S, Harrison J, Spalding M, Lysenko I (2005) Measuring the extent and 21. Cardoso P, Erwin TL, Borges PAV, New TR (2011) The seven impediments in effectiveness of protected areas as an indicator for meeting global biodiversity invertebrate conservation and how to overcome them. Biol Conserv 144: targets. Phil Trans Roy Soc Lond Ser B 360: 443–455. 2647–2655. 3. Ricketts TH, Daily GC, Ehrlich PR, Fay JP (2001) Countryside biogeography of 22. Dunn RR (2005) Modern insect extinctions, the neglected majority. Conserv moths in a fragmented lands cape: biodiversity in native and agricultural Biol 19: 1030–1036. habitats. Conserv Biol 15: 378–388. 23. Lomolino MV (2004) Conservation biogeography. In: Lomolino MV, 4. Hughes JB, Daily GC, Ehrlich PR (2002) Conservation of tropical forest birds in Heaney LR, eds. Frontiers of Biogeography: New Directions in the Geography countryside habitats. Ecol Lett 5: 121–129. of Nature. Sunderland: Sinauer Associates. pp 293–296. 5. Bhagwat SA, Willis KJ, Birks HJB, Whittaker RJ (2008) Agroforestry: A refuge 24. Mokany K, Ferrier S (2011) Predicting impacts of climate change on for tropical biodiversity? TREE 23: 261–267. biodiversity: a role for semi-mechanistic community-level modelling. Divers 6. Franklin JF, Lindenmayer DB (2009) Importance of matrix habitats in Distrib 17: 374–380. maintaining biological diversity. Proc Natl Acad Sci USA 106: 359–350. 25. Elith J, Graham CH, Anderson RP, Dudik M, Ferrier S, et al. (2006) Novel 7. Margules C, Sarkar S (2007) Systematic Conservation Planning. Cambridge: methods improve prediction of species’ distributions from occurrence data. Cambridge University Press. 270 p. Ecography 29: 129–151. 8. Moilanen A, Wilson KA, Possingham HP (2009) Spatial Conservation 26. Elith J, Leathwick J (2009) The contribution of species distribution modelling to Prioritization. Oxford: Oxford University Press. 320 p. conservation prioritization. In: Moilanen A, Wilson KA, Possingham HP, eds. 9. Rodrigues ASL, Tratt R, Wheeler BD, Gaston KJ (1999) The performance of Spatial conservation prioritization. Oxford: Oxford University Press. pp 70–93. existing networks of conservation areas in representing biodiversity. Proc Roy 27. Martıń JL, Cardoso P, Arechavaleta M, Borges PAV, Faria BF, et al. (2010) Soc Lond Ser B 266: 1453–1460. Using taxonomically unbiased criteria to prioritize resource allocation for oceanic island species conservation. Biodivers Conserv 19: 1659–1682. 10. Margules C, Pressey RL (2000) Systematic conservation planning. Nature 405: 28. Cardoso P, Borges PAV, Triantis K, Ferrańdez MA, Martıń JL (2011) Adapting 243–253. the IUCN Red List criteria for invertebrates. Biol Conserv 144: 2432–2440. 11. Sarkar S, Pressey RL, Faith DL, Margules CR, Fuller T, et al. (2006) 29. Borges PAV, Costa A, Cunha R, Gabriel R, Gonçalves V, et al. (2010) A List of Biodiversity Conservation Planning Tools: Present Status and Challenges for the the Terrestrial and Marine Biota from the Azores. Cascais: Princı´pia. 432 p. future. Ann Rev Environ Res 31: 123–59. 30. Ramos JA (1995) The diet of the Azores bullfinch Pyrrhula murina and floristic 12. Pressey RL, Cabeza M, Watts ME, Cowling RM, Wilson KA (2007) variation within its range. Biol Conserv 71: 237–249. Conservation planning in a changing world. TREE 22: 583–592. 31. Dias E, Elias R, Nunes V (2004) Vegetation mapping and nature conservation: a 13. Arponen A, Moilonen A, Ferrier S (2008) A successful community-level strategy case study in Terceira Island (Azores). Biodivers Conserv 13: 1519–1539. for conservation prioritization. J Appl Ecol 45: 1436–1445. 32. Borges PAV, Aguiar C, Amaral J, Amorim IR, Andre´ G, et al. (2005) Ranking 14. De Vries MFW, Poschlod P, Willems JH (2002) Challenges for the conservation protected areas in the Azores using standardized sampling of soil epigean of calcareous grasslands in northwestern Europe: integrating the requirements of arthropods. Biodivers Conserv 14: 2029–2060. flora and fauna. Biol Conserv 104: 265–273. 33. Borges PAV, Ugland KI, Dinis FO, Gaspar C (2008) Insect and spider rarity in 15. Van Solomon M, Jaarsveld AS, Biggs HC, Knight M (2003) Conservation an oceanic island (Terceira, Azores): true rare and pseudo-rare species. In: targets for viable species assemblages? Biodivers Conserv 12: 2435–2441. Fattorini S, ed. Insect Ecology and Conservation. Kerala: Research Signpost. pp 16. Zhu H, Qin P, Wang H (2004) Functional group classification and target species 47–70. selection for Yancheng Nature Reserve, China. Biodivers Conserv 13: 34. Cardoso P, Borges PAV, Gaspar C (2007) Biotic integrity of the arthropod 1335–1353. communities in the natural forests of Azores. Biodivers Conserv 16: 2883–2901. 17. Dennis RLH, Shreeve TG, Van Dyck H (2006) Habitats and resource-based 35. Cardoso P, Lobo JM, Aranda SC, Dinis F, Gaspar C, et al. (2009) A spatial scale definition to conserve butterflies. Biodivers Conserv 15: 1943–1966. assessment of habitat effects on arthropod communities of an oceanic island. 18. Sarkar S (1999) Wilderness preservation and biodiversity conservation—keeping Acta Oecol 35: 590–597. divergent goals distinct. Bio Science 49: 405–412. 36. Gaspar C, Gaston KJ, Borges PAV, Cardoso P (2011) Selection of priority areas 19. New TR (2011) Strategic planning for invertebrate species conservation - how for arthropod conservation in the Azores archipelago. J Insect Conserv 15: effective is it? J Threat Taxa 3: 2033–2044. 671–684.

PLoS ONE | www.plosone.org 8 March 2012 | Volume 7 | Issue 3 | e33995 Arthropod Rarity and Hotspot Identification

37. Mcdonald RI, Boucher TM (2011) Global development and the future of the 56. Fattorini S (2006a) A new method to identify important conservation areas protected area strategy. Biol Conserv 144: 383–392. applied to the butterflies of the Aegean Islands (Greece). Anim Conserv 9: 38. Gaspar C, Borges PAV, Gaston KJ (2008) Diversity and distribution of 75–83. arthropods in native forests of the Azores archipelago. Arquipe´lago Life and 57. Leroy B, Petillon J, Gallon R, Canard A, Ysnel F (2012) Improving occurrence- marine Sciences 25: 1–30. based rarity metrics in conservation studies by including multiple rarity cut-off 39. Meijer SS, Whittaker RJ, Borges PAV (2011) The effects of land-use change on points. Insect Conserv DiverDOI: 10.1111/j.1752-4598.2011.00148.x. In press. arthropod richness and abundance on Santa Maria Island (Azores): unmanaged 58. Fattorini S (2009) Assessing priority areas by imperilled species: insights from the plantations favour endemic beetles. J Insect Conserv 15: 505–522. European butterflies. Anim Conserv 12: 313–320. 40. Hartley S, Kunin W (2003) Scale dependency of rarity extinction risk and 59. Dapporto L, Dennis RLH (2008) Island size is not the only consideration. conservation priority. Conserv Biol 17: 1559–1570. Ranking priorities for the conservation of butterflies on Italian offshore islands. 41. Rabinowitz DS (1981) Seven forms of rarity. In: Synge H, ed. The Biological J Insect Conserv 12: 237–249. aspects of rare plant conservation. Chichester: Wiley. pp 205–217. 60. Bergerot B, Lasne E, Vigneron T, Laffaille P (2008) Prioritization of fish 42. Rabinowitz D, Cairns S, Dillon T (1986) Seven forms of rarity and their assemblages with a view to conservation and restoration on a large scale frequency in the flora of the British Isles. In: Soule´ ME, ed. Conservation European basin the Loire (France). Biodivers Conserv 17: 2247–2262. Biology: the science of scarcity and diversity. Sunderland: Sinauer Associates. pp 61. Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of 182–204. species geographic distributions. Ecol Model 190: 231–259. 43. Kattan G (1992) Rarity and vulnerability: the birds of the Cordillera Central of 62. Elith J, Phillips SJ, Hastie T, Dudı´k M, Chee YE, et al. (2011) A statistical Colombia. Conserv Biol 6: 64–70. explanation of MaxEnt for ecologists. Divers Distrib 17: 43–57. 44. Dobson FS, Yu J (1993) Rarity in Neotropical forest mammals revised. Conserv 63. Borges PAV, Lobo JM, Azevedo EB, Gaspar C, Melo C, et al. (2006) Invasibility Biol 7: 586–591. and species richness of island endemic arthropods: a general model of endemic vs. exotic species. J Biogeogr 33: 169–187. 45. Manne LL, Pimm SL (2001) Beyond eight forms of rarity: which species are 64. Azevedo EB, Pereira LS, Itier B (1999) Modelling the local climate in island threatened and which will be next? Anim Conserv 4: 221–229. environments: water balance applications. Agr Water Manage 40: 393–403. 46. Fattorini S (2008) A multidimensional characterization of rarity applied to the 65. DROTRH (2007) Carta de ocupaçaõ do solo da regiaõ Autońoma dos Açores - Aegean tenebrionid beetles (Coleoptera Tenebrionidae). J Insect Conserv 12: Projecto SUEMAC. Ponta Delgada: Secretaria Regional do Ambiente, Direcçaõ 251–263. Regional do Ordenamento do territo´rio e dos Recursos Hı´dricos. 57 p. 47. Fattorini S (2010) Use of insect rarity for biotope prioritisation: the tenebrionid 66. Pineda E, Lobo JM (2009) Assessing the accuracy of species distribution models beetles of the Central Apennines (Italy). J Insect Conserv 14: 367–378. to predict amphibian species richness patterns. J Anim Ecol 78: 182–190. 48. Fattorini S (2010) Biotope prioritisation in the Central Apennines (Italy): species 67. Liu C, Berry PM, Dawson TP, Pearson RG (2005) Selecting thresholds of rarity and cross-taxon congruence. Biodivers Conserv 19: 3413–3429. occurrence in the prediction of species distributions. Ecography 28: 385–393. 49. Fattorini S (2011) Insect rarity, extinction and conservation in urban Rome 68. Brown JH (1995) Macroecology. Chicago and London: The University of (Italy): a 120-year-long study of tenebrionid beetles. Insect Conserv Diver 4: Chicago Press. 284 p. 307–315. 69. Gaston KJ, Borges PAV, He F, Gaspar C (2006) Abundance, spatial variance 50. Gabriel R, Homem N, Couto A, Aranda SC, Borges PAV (2011) Azorean and occupancy: arthropod species distribution in the Azores. J Anim Ecol 75: Bryophytes: a preliminary review of rarity patterns. In: Martins AMF, 646–656. Carvalho MC, eds. Celebrating Darwin: Proceedings of the Symposium 70. Koomen P, van Helsdingen PJ (1996) Listing of biotopes in Europe according to ‘‘Darwin’s Mistake and what we are doing to correct it’’, Ponta Delgada, 19– their significance for invertebrates. Nature and Environmnent, no. 77. 22 September, 2009, Açoreana, Supl. 7. pp 149–206. Strasbourg: Council of Europe Publishing. 74 p. 51. Gaston KJ (1994) Rarity. London: Chapman & Hall. 205 p. 71. Primack RB (2006) Essentials of conservation biology, 4th edn. Sunderland: 52. Gaston KJ, Blackburn TM (2000) Pattern and process in macroecology. Oxford: Sinauer Associates Inc Publishers. 538 p. Blackwell Science. 392 p. 72. New TR (2008) Insect species conservation. Cambridge: Cambridge University 53. Myers AA, De Grave S (2000) Endemism: origins and implications. Vie Milieu Press. 272 p. 50: 195–204. 73. Cardoso P (2011) Diversity and community assembly patterns of epigean vs. 54. Cook JA, MacDonald SO (2001) Should endemism be a focus of conservation troglobiont spiders in the Iberian Peninsula. Int J Speleol 41: 83–94. efforts along the North Pacific Coast of North America? Biol Conserv 97: 74. Hopkins MJG (2007) Modelling the known and unknown plant biodiversity of 207–213. the Amazon Basin. J Biogeogr 34: 1400–1411. 55. Devictor V, Clavel J, Julliard R, Lavergne S, Mouillot D, et al. (2010) Defining 75. Fattorini S (2006) Detecting biodiversity hotspots by species–area relationships: a and measuring ecological specialization. J Appl Ecol 47: 15–25. case study of Mediterranean beetles. Conserv Biol 20: 1169–1180.

PLoS ONE | www.plosone.org 9 March 2012 | Volume 7 | Issue 3 | e33995